Running and troubleshooting

Sequences end. When one finishes, sending stops, replies taper off soon after, and nothing flags it: a finished sequence looks just like a healthy one. You notice weeks later when the pipeline is empty. This is the most common way a new outbound program dies, and the fix is one habit.

The rule

Keep two campaigns built and ready behind the one that’s running. When a live sequence is nearly done, the next one should already be written, tested and published, ready to activate.

The weekly hour

One standing hour, same slot every week.

1. Read the numbers on what’s live

Open each sequence to see its sends, replies and reply rate, and the Replies tab for what people said.

• Reply rate matters most. Cold reply rates are low, and around 1% isn’t unusual. Warm plays, such as reviving a lost deal or writing to a past champion, run far higher. These are rough industry figures, not targets.
• Interested replies are the share who genuinely want to talk. Expect about half of all replies to be out-of-office messages, “not interested” or “wrong person”. That’s normal.
• Bounce rate is a health measure, not a performance one. Keep it under 1%. See Sending safely.
• Open rate is close to meaningless, because some mail clients register an open whether or not anyone read the email. Its one use: a very low open rate (under about 30%) usually means your mail is going to spam.

What the numbers tell you to fix

What you see	What’s actually wrong
Very low open rate	Mail isn’t reaching inboxes. It’s a sending problem, not a writing problem. Check your domain authentication, then whether addresses were verified
Opens look fine, replies very low	The message arrives and fails. Either the copy is weak or you’re writing to the wrong people
Plenty of replies, few interested	You’re reaching the wrong people. Tighten targeting rather than rewriting
Strong response to the first email, nothing after	Follow-ups aren’t adding anything. Each needs a new angle, not a reminder
Replies fall off sharply partway through	Deliverability is degrading. Check bounces per sender
Interested replies, nothing books	The ask is too big, or you’re reaching people who are interested but can’t buy

Work out whether it’s a delivery, targeting or writing problem before you change anything. Most people rewrite the copy when their mail was going to spam.

2. Decide the next play

A new cohort, a new signal, a new segment, or a fresh angle on the same segment. If you’re stuck, ask the agent: “Show me three plays we haven’t tried yet that make sense for this business.” Work down the Outbound playbook rather than reaching for cold every time.

3. Build and test it so it’s queued

Queued, not activated: built, test-run, published, and ready to activate the day the current one winds down.

Where the next play comes from

• Your own results. A cohort that replied well deserves a second angle. One that flatlined deserves a post-mortem, not another attempt.
• Your CRM, again. New lost deals pile up, people change jobs, new stakeholders appear at customers. Put revive, expand and champions-who-moved plays on a repeating schedule.
• New signals. A signal you weren’t tracking three months ago is a new play against the same accounts.
• Your business. A launch, a case study, a conference or a feature that closes a gap you used to lose on.

When something about your targeting or messaging changes, change it in the knowledge base rather than patching individual sequences. See Preparing to send. Refresh your ICP quarterly, or whenever you add around ten closed-won deals.

Re-verify a list before you reuse it. Contact data decays constantly, and a list that was clean in March will bounce in September.

Weekly checklist

• Numbers read on everything live, per sender as well as per campaign
• Any sender with rising bounces paused and investigated
• LinkedIn acceptance rate checked
• Next play decided
• Next sequence built, tested and published, not activated
• Knowledge base updated if you learned something worth keeping

Troubleshooting

If something here doesn’t match what you see, ask the agent in chat. It can read your workspace’s sequences, enrollments and settings, which is usually faster than working through a list. Describe what you expected and what happened.

Setup problems

“Step 7 says it can’t run.” The Find contacts for account automation from step 6 is still a draft. Test it and activate it. This is the most common setup blocker, and the message points at step 7, not the cause. See Preparing to send.

“Step 5 won’t score anything.” The same problem one step earlier: the Score prospect account automation from step 4 is still a draft.

“It says my ICP files are missing.” Steps 3 onward read the knowledge base. Run build my ICP and build my scoring rubric first.

“I can’t create the automations.” Creating automations needs a workspace admin. Ask an admin to run steps 4 and 6.

Scoring and list problems

“Scores are all clustered in the same range.” Your rubric isn’t telling companies apart. Fix the rubric and re-score rather than adjusting individual accounts. The fix is usually more weight on fewer dimensions.

“The scores don’t match my judgment.” Read the rationale on a few accounts. If it leans on a field you don’t trust, that field is the problem. If it reasons correctly from a premise you disagree with, your ICP is the problem.

“Some of my accounts came back unscored.” When you find contacts, the agent shows unscored accounts separately and asks whether to score them, include them or skip them. Nothing is silently cut.

“My list has poor email coverage.” Run a verification pass, drop what you can’t reach, and consider LinkedIn for the rest. Never enroll a list you haven’t verified.

“Verification passed but I’m still bouncing.” Catch-all domains accept any address, so a check can’t confirm a specific mailbox. By default Lightfield removes those recipients from sequences unless Send to unverifiable emails is turned on in Settings → Sequences.

“The agent screened out accounts I would have kept.” Your anti-icp.md is too broad. Because it also shapes the low end of your scoring bands, an over-aggressive anti-ICP suppresses accounts twice.

Sequence problems

“My sequence won’t activate and it mentions LinkedIn.” No healthy LinkedIn account is connected and designated on the sequence. In Settings → Sequences, open your sender page and connect LinkedIn, then make sure the account is designated on the sequence. If the message says your plan doesn’t include LinkedIn, the LinkedIn steps need a plan that includes LinkedIn sequences.

“My sequence won’t activate and it mentions a sending mailbox.” No healthy email sender is designated on the sequence. Turn on Enable for sequences for your connected account in Settings → Sequences, or fix the mailbox it names, then designate it on the sequence.

“A test run failed saying no sender is available.” Ask the agent to check which senders are designated on the sequence and whether Use connected account is set the way you intend. If you only send from your connected account, tell the agent so. Don’t publish just to try to fix this: test runs use your current draft and never need a publish.

“Sending stopped and nothing looks broken.” Check for a paused enrollment first. If a message can’t pass the pre-send quality checks (an em dash is the usual cause), the step fails and that contact’s enrollment pauses. Paused enrollments don’t restart on their own. Ask the agent to find them and resume them. Also check whether a sender shows Paused after bounces, or whether a LinkedIn limit paused the sequence.

“I enrolled contacts and nothing sent.” The sequence hasn’t been activated. Publishing doesn’t start sending; Activate does. The first publish leaves a sequence paused. Also check the send window: sends outside it wait until it opens.

“I edited the copy but the old version is still going out.” Publish the change. The agent only publishes when you ask it to, so confirm your last edit was published. You don’t need to pause the sequence to edit it.

“My LinkedIn step failed on contacts I know.” It isn’t because you’re already connected: an invite to an existing connection succeeds without sending a new request. Keep the invite step (Lightfield rejects DM-only sequences) and ask the agent to check the enrollment for the actual error.

“I asked for bold text or a hyperlink and it didn’t work.” Sequence emails are plain text by design, and markdown or HTML formatting is blocked. Paste the raw URL.

“I want to rename something across all my sequences at once.” Edits that keep a sequence’s existing steps apply to enrolled contacts at their next step, with no need to pause. Edits that replace, remove or reorder steps can restart contacts or change their next step, and pausing doesn’t prevent that. Edit one sequence at a time and read the impact the agent reports before you publish.

“I want to change the gap between two steps.” Ask the agent. Timing is changed in chat, not in a field.

“The agent asked me questions I can’t answer.” You don’t need every answer. Say which list you’re targeting, say you’re unsure about the rest, and ask for options. Reacting to three proposals is easier than answering from scratch.

“The copy sounds like AI.” Two fixes, in order: give the agent a concrete list of phrases to avoid, and paste in two of your own emails as a voice reference. Concrete bans and real examples do far more than “make it sound human”. See Building sequences.

“Every preview has the same opener with different names.” The personalization isn’t finding real signals. Either your contacts have thin data, or the sequence doesn’t tell the agent to find a signal per contact with a fallback line. Run test runs across several cohorts to see how widespread it is.

Performance problems

“Replies dropped off and I don’t know why.” Check whether your sequences finished. A flat pipeline a few weeks after a sequence completes is the most common cause, and the easiest to miss.

“Open rate is low.” A very low open rate is a deliverability problem, not a subject-line one. Check domain authentication, then verification.

“Good reply rate, nothing books.” Either the ask is too big or you’re reaching people who can’t buy. “Sounds interesting, but I’d need to loop in X” is a targeting problem. Polite, non-committal replies usually mean the ask was too heavy.

“Most of my replies are people saying no.” Normal. What matters is the share who are genuinely interested.

“Bounce rate spiked mid-campaign.” Pause the affected sender, not the whole campaign. Isolate the failing segment, re-verify it, and resume from a healthy sender. See Sending safely.

“LinkedIn stopped letting me send requests.” The account hit LinkedIn’s invitation limit. Lightfield pauses the sequences sending from that account and notifies your admins. Resume them after about seven days. If it happened early, your acceptance rate may have dropped and LinkedIn tightened your allowance.

Known limits

These are by design, not bugs, and there’s no setting to change them:

• Sequences run in a straight line, with no conditional branching.
• Email and LinkedIn only. No SMS or dialer.
• Step timing and content are changed by talking to the agent.
• Lists don’t refresh themselves as signals appear. Re-run them.
• Overnight send windows aren’t supported.
• They’re not built for a rep sending thousands of emails a week through rotating inboxes.

What sequences are built for: reaching a few hundred specific people, each with a message that reflects something real about them.

Next steps

• Outbound playbook: pick the next play.
• Building sequences: build it.
• Sending safely: limits, checks and deliverability.